153
6.0 VOLUMETRIC ESTIMATION Keywordsdeterministic methods, STOIIP, GIIP, reserves, ultimate recovery, net oil sands, area-depth and area-thickness methods, gross rock volume, expectation curves, probability of excedence curves, uncertainty, probability of success, annual reporting requirements, Monte-Carlo simulation, parametric method lnfroducfion and Commercia/ Application: Volumetric estimation is concerned with quantifying how much oil and gas exists in an accumulation. The estimate will vary throughout the field lifetime as more information becomes available and as the technology for gathering and interpreting the data improves. A volumetric estimate is therefore a current estimate, and should be expected to change over time. Two main methods of estimating volumetrics are used; deterministic and probabilistic. Deterministic methods average the data gathered at points in the reservoir, from well logs, cores, seismic, to estimate the field-wide properties. Probabilistic methods use predictive tools, statistics, analogue field data and input regarding the geological model to predict trends in reservoir properties away from the sample points. This section will concentrate on the deterministic methods and the techniques used for expressing uncertainty in these volumetric estimates. The volumetrics of a field, along with the anticipated recovery factors, control the reserves in the field; those hydrocarbons which will be produced in the future. The value of an oil or gas company lies predominantly in its hydrocarbon reserves which are used by shareholders and investors as one indication of the strength of the company, both at present and in the future. A reliable estimate of the reserves of a company is therefore important to the current value as well as the longer term prospects of an oil or gas company.
6.1 Deterministic Methods
Volumetric estimates are required at all stages of the field life cycle. In many instances a first estimate of “how big” an accumulation could be is requested. If only a “back of the envelope” estimate is needed or if the data available is very sparse a quick look estimation can be made using field wide averages.
154
The formulae to calculate volumes of oil or gas are STOllP = GRV
.
N
- . G
GllP = GRV
.
N
- .
1
so
$j
4
. ; s
‘
- [stb] Bo 1
- [scf]
G
B,
Ultimate recovery = HCllP * Recovery factor
[stb] or [scf]
Reserves = UR - Cumulative Production
[stb] or [scf]
“ST0IIP”is a term which normalises volumes of oil contained under high pressure and temperature in the subsurface to surface conditions (e.g. 1 bar, 15°C). In the early days of the industry this surface volume was referred to as “stocktankoil’ and since measured prior to any production having taken place it was the volume “initially in place”. “GIIP‘ is the equivalent expression for gas initially in place. Ultimate recovery (UR) and reserves are linked to the volumes initially in place by the recovery factor, or fraction of the in place volume which will be produced. Before production starts reserves and ultimate recovery are the same.
“GRV’ is the gross rock volume of the hydrocarbon-bearing interval and is the product of the area (A) containing hydrocarbons and the interval thickness (H), hence
[VI or [acre.ft] or [m3]
GRV=A. H
The area can be measured from a map. Figure 6.1 clarifies some of the reservoir definitions used in reserves estimation. Well 1
Well 2
1 Net sand Net oil sand = Gross interval thickness
I
Figure 6.1 Definitions used for volumetric estimates
155
H
=
the isochore thickness of the total interval (“gross thickness”) regardless of lithology.
Net sand
=
the height of the lithologic column with reservoir quality, i.e. the column that can potentially store hydrocarbons.
Net Oil Sand (NOS) =
the length of the net sand column that is oil bearing.
The other parameters used in the calculation of STOllP and GllP have been discussed in Section 5.4 (Data Interpretation). The formation volume factors (Bo and Bg) were introduced in Section 5.2 (Reservoir Fluids). We can therefore proceed to the “quick and easy” deterministic method most frequently used to obtain a volumetric estimate. It can be done on paper or by using available software. The latter is only reliable if the software is constrained by the geological reservoir model.
-
6.1.1 The area depth method
From a top reservoir map (Fig. 6.2) the area within a selected depth interval is measured. This is done using a planimeter, a hand operated device that measures areas.
Depth ’
~
~
1000
1200 1400
Area C H = Thickness of reservoir from logs
Figure 6.2 The area - depth method
The stylus of the planimeter is guided around the depth to be measured and the respective area contained within this contour can then be read off. The area is now plotted for each depth as shown in Figure 6.2 and entered onto the area - depth graph. Since the structure is basically cut into slices of increasing depth the area measured for each depth will also increase. Connecting the measured points will result in a curve describing the area - depth relationship of the top of the reservoir. If we know the gross thickness (H) from logs we can establish a second curve representing the area - depth plot for the base of the reservoir. The area between the two lines will equal the volume of rock between the two markers. The area above the OWC is the oil bearing GRV. The other parameters to calculate STOllP can be taken as averages from our petrophysical evaluation (see Section 5.4.). Note that this method assumes that the reservoir thickness is constant across the whole field. If this is not a reasonable approximation, then the method is not applicable, and an alternative such as the area -thickness method must be used (see below). This procedure can be easily carried out for a set of reservoirs or separate reservoir blocks. It is especially practical if stacked reservoirs with common contacts are to be evaluated. In cases where parameters vary across the field we could divide the area into sub blocks of equal values which we measure and calculate separately.
-
6.1.2 The area thickness method
In some depositional environments, e.g. fluviatile channels, marked differences in reservoir thickness will be encountered. Hence the assumption of a constant thickness, or a linear trend in thickness across the field will no longer apply. In those cases a set of additional maps will be required. Usually a net oil sand (NOS) map will be prepared by the production geologist and then used to evaluate the hydrocarbon volume in place.
In the following example, well 1 has found an oil bearing interval in a structure (1). An OWC was established from logs and has been extrapolated across the structure assuming continuous sand development. However, the core (in reality cores from a number of wells) and 3D seismic have identified a channel depositional environment. The channel has been mapped using specific field data and possibly analogue data from similar fields resulting in a net sand map (2). In this case the hydrocarbon volume is constrainedby the structural feature of the field andthe distribution of reservoir rock i.e. the channel geometry. Hence we need to combine the two maps to arrive at a net oil sand map (3). The “odd shape” is a result of that combination and actually it is easy to visualise: at the fault the thickness of oil bearing sand will rapidly decrease to zero. The same is the case at the OWC. Where the net sand map indicates 0 m there will be 0 m of net oil sand. Where the channel is best developed showing maximum thickness we will encounter the maximum net oil sand thickness, but only until the channel cuts through the fault or the OWC.
157
We can now planimeter the thickness of the different NOS contours, plot thickness versus area and then integrate both with the planimeter. The resulting value is the volume of net oil sand (4) and not the GRV!
It is clear that if the area - depth method had been applied to the above example, it would have led to a gross over-estimationof STOIIP. It would also have been impossible to target the best developed reservoir area with the next development well.
+
Net sand map (m) (channel)
+
0 Net oil sand volume
Net oil sand (m)
Figure 6.3 Net oil sand mapping and area -thickness method
158
It should be noted that our example used a very simple reservoir model to show the principle. NOS mapping is usually a fairly complex undertaking.
As will be shown in the next section, the methods discussed so far do not take account of the uncertainties and lateral variations in reservoir parameters. Hence the accuracy of the results is not adequate for decision making. The next section introduces a more comprehensive approach to volumetric estimation.
6.2 Expressing uncertainty
As shown in the Section 6.1, the calculation of volumetrics for a field involves the combination of a number of input parameters. It should be realised that each of these has a range of uncertainty in its estimation. The extent of this range of uncertainty will depend upon the amount of data available, and the accuracy of that data. The value in combining ranges of uncertainty in the input parameters to give a range of estimates for STOIIP, GllP and UR, is that both upside potential and downside risks can be quantified. Using a single figure to represent, say STOIIP, may lead to missed opportunities, or unrecognised risk. The range of uncertainty in the UR may be too large to commit to a particular development plan, and field appraisal may be required to reduce the uncertainty and allow a more suitable development plan to be formed. Unless the range of uncertainty is quantified using statistical techniques and representations, the need for appraisal cannot be determined. Statistical methods are used to express ranges of values of STOIIP, GIIP, UR, and reserves.
6.2.1 The input to volumetric estimates
The input parameters to the calculation of volumetrics were introduced at the beginning of Section 6.1. Let us take the STOllP calculation as an example. STOllP = GRV .
N -
G
.
@
'
s o .
1
- [stb] BO
Each of the input parameters has an uncertainty associated with it. This uncertainty arises from the inaccuracy in the measured data, plus the uncertainty as to what the values are for the parts of the field for which there are no measurements. Take for example a field with five appraisal wells, with the following values of average porosity for a particular sand:
159
Figure 6.4 Porosity distribution in a field It would be unrealistic to represent the porosity of the sand as the arithmetic average of the measured values (0.20), since this would ignore the range of measured values, the volumes which each of the measurements may be assumed to represent, and the possibility that the porosity may move outside the range away from the control points. There appears to be a trend of decreasing porosity to the south-east, and the end points of the range may be 0.25 and 0.15, Le. larger than the range of measurements made. An understanding of the geological environment of deposition and knowledge of any diagenetic effects would be required to support this hypothesis, but it could only be proven by further data gathering in the extremities of the field. When providing input for the STOllP calculation a range of values of porosity (and all of the other input parameters) should be provided, based on the measured data and estimates of how the parameters may vary away from the control points. The uncertainty associated with each parameter may be expressed in terms of a probability density function, and these may be combined to create a probability density function for STOIIP. It is common practice within oil companies to use expectation curves to express ranges of uncertainty. The relationship between probability density functions and expectation curves is a simple one.
6.2.2 Probability density functions and expectation curves
A well recognised form of expressing uncertainty is the probability density function. For example, if one measured the heights of a class of students and plotted them on a histogram of height ranges against the number of people within that height range, one might expect a relative frequency distribution plot, also known as a probability density function (PDF) with discrete values, such as that in Figure 6.5. Each person measured is represented by one square, and the squares are placed in the appropriate height category. If the value on the x-axis were continuous rather than split into discrete ranges, the discrete PDF could be represented by a continuous function. This is useful in predicting
160
what fraction of the population have property X (height in our example) greater than a chosen value (X,). From the continuous PDF one would estimate that approximately 70% of the population sampled were of height greater than or equal to XI. In other words, if one were to randomly pick a person from the sample population, there is a 70% probability that the height of that person is greater than or equal to height XI. There is a 100% probability that the person is greater than or equal to height Xmi,, and a 0% chance that the person is greater than height.,,X, The expectation curve is simply a representation of the cumulative probability density function: freq ( 4
Discrete Values
X
min
X
X
1
Figure 6.5 A probability density function
max
161
fregx)
Probability Density Function
X I
““7
:\
I I I I I
0
X
X
min
11
I
\!
x
Expectaition Curve
I
I I I I I I I I I
max
X
1 X
Figure 6.6 The probability density function and the expectation curve For oilfield use, the x-axis on expectation curves is typically the STOIIP, GIIP, UR, or Reserves of a field. Expectation curves are alternatively known as ‘probability of excedence curves’ or ‘reverse cumulative probability curves’. This text will use the term ‘expectation curve’ for conciseness. The slope of the expectation curve indicates the range of uncertainty in the parameter presented: a broad expectation curve represents a large range of uncertainty, and a steep expectation curve represents a field with little uncertainty (typical of fields which have much appraisal data, or production history).
162
0
100
200 300 STOllP (MMstb)
400
Figure 6.7 Types of expectation curve Expectation Curves 1 and 2 represent discoveries, since they both have a 100% probability of containing a finite amount of oil (greater than zero). Case 1 is a well defined discovery since the range of uncertainty in STOllP values is small (at least 100 MMstb, but less than 220 MMstb). By contrast, case 2 represents a poorly defined discovery, with a much broader range of STOIIP, and would probably require appraisal activity to reduce this range of uncertainty before committing to a development plan. Cases 3 and 4 are both exploration prospects, since the volumes of potential oil present are multiplied by a chance factor which represents the probability of there being oil there at all. For example, case 3 has an estimated probability of oil present of 65%, i.e. low risk of failure to find oil (35%). However, even if there is oil present, the volume is small; no greater than 130 MMstb. This would be a low risk, low reward prospect. Case 4 has a high risk of failure (85%) to find any oil, but if there is oil there then the volume in place might be quite large (up to 400 MMstb). This would class as a high risk, high reward prospect. Expectation Curves for a Discovery For a discovery, a typical expectation curve for Ultimate Recovery is shown in figure 6.8.
163
Figure 6.8 Expectation curve for a discovery For convenience, the probability axis may be split into three equal sectors in order to be able to represent the curve by just three points. Each point represents the average value of reserves within the sector. Again for convenience, the three values correspond to chosen cumulative probabilities (85%, 50%, and 15%), and are denoted by the values: Low estimate
=
85% cumulative probability (i.e. 85% probability of at least these reserves)
Medium estimate
=
50% cumulative probability
High estimate
=
15% cumulative probability
The percentages chosen are often denoted as the p85, p50, p l 5 values. Because they each approximately represent one third of the distribution, their discrete probabilities may each be assigned as one third. This approximation is true for a normal (or symmetrical) PDF. If the whole range is to be represented by just one value (which of course gives no indication of the range of uncertainty), then the "expectation value " is used: Expectation Value
High + Medium + Low =
3 An alternative and commonly used representation of the range of reserves is the proven, proven plus probable, and proven plus probable plus possible definition. The exact cumulative probability which these definitions correspond to on the expectation curve
164
for Ultimate Recovery varies from country to country, and sometimes from company to company. However, it is always true that the values lie within the following ranges proven
:
between 100% and 66%
proven + probable
:
between 66% and 33%
proven + probable + possible
:
between 33% and 0%
The annual reporting requirements to the US Securities and Exchange Commission (SEC) legally oblige listed oil companies to state their proven reserves. Many companies choose to represent a continuous distribution with discrete values using the p90, p50, p l 0 values. The discrete probabilities which are then attached to these values are then 25%, 50%, 25%, for a normal distribution. Expectation Curves for an Exploration Prospect When an explorationist constructs an expectation curve, the above approach for the volumetrics of an accumulation is taken, but one important additional parameter must be taken into account : the probability of there being hydrocarbons present at all. This probability is termed the “ProbabilityofSuccess”(POS),and is estimated by multiplying together the probability of there being: a source rock where hydrocarbons were generated a reservoir in a structure which may trap hydrocarbons a seal on top of the structure to stop the hydrocarbons migrating further a migration path for the hydrocarbons from source rock to trap the correct sequence of events in time (trap present as hydrocarbons migrated) The estimated probabilities of each of these events occurring are multiplied together to estimate the POS, since they must alloccur simultaneously if a hydrocarbon accumulation is to be formed. If the POS is estimated at say 30%, then the probability of failure must be 70%, and the expectation curve for an exploration prospect may look as shown in Figure 6.9. As for the expectation curve for discoveries, the “success” part of the probability axis can be divided into three equal sections, and the average reserves for each section calculated to provide a low, medium and high estimate of reserves, if there are hydrocarbons present.
165
100
-
\
P (x) Yo 80 -
> Probability of Failure = 70%
60 -
40
-
Figure 6.9 Expectation curve for an exploration prospect 6.2.3 Generating expectation curves
Returning to the input parameters for an ultimate recovery calculation, we have established that UR
=
GRV .
N
- .
1
$
.
s o .
G
- .
RF [stb]
BO
Each of the input parameters requires an estimate of the range of values, which can itself be represented by a distribution, or expectation curve. Ideally, the expectation curves for the input parameters are combined together statistically. Some variables often have dependencies, such as reservoir porosity and permeability (a positive correlation) or the capital cost of a specific equipment item and its lifetime maintenance cost (a negative correlation). We can test the linear dependency of two variables (say x and y) by calculating the covariance between the two variables (oxy) and the correlation coefficient (r): 1 i=n ox,,= - ~ ( x i - p x ) * ( y i - p y )and
n
i=l
(5
r =xy oxoy
The value of r varies between plus and minus one, the positive values indicating a positive correlation (as x increases, so does y), and the negative values indicating a
166
negative correlation (as x increases, y decreases). The closer the absolute value of r is to 1 .O, the stronger the correlation. A value of r = 0 indicates that the variables are unrelated. Once we are satisfied that a true dependency exists between variables, we can generate equations which link the two using methods such as the least squares fit technique. If a correlation coefficient of 1.O were found, it would make more sense to represent the relationship in a single line entry in the economic model. There is always value in cross-plotting the data for the two variables to inspect the credibility of a correlation. As a rough guide, correlation factors above 0.9 would suggest good correlation I
6.2.4 The Monte Carlo Method
This is the method used by the commercial software packages “Crystal Ball” and “@RISK. The method is ideally suited to computers as the description of the method will reveal. Suppose we are trying to combine two independent variables, say gross reservoir thickness and net-to-gross ratio (the ratio of the net sand thickness to the gross thickness of the reservoir section) which need to be multiplied to produce a net sand thickness. We have described the two variables as follows:
Figure 6.10 Probability distributions for two variables; input for Monte Carlo A random number (between 0 and 1) is picked, and the associated value of gross reservoir thickness (T) is read from within the range described by the above distribution. The value of T close to the mean will be randomly sampled more frequently than those values away from the mean. The same process is repeated (using a different random number) for the net-to-gross ratio (N/G). The two values are multiplied to obtain one value of net sand thickness. This is repeated some 1,000-10,000 times, with each outcome being equally likely. The outcomes are used to generate a distribution of values of net sand thickness. This can be performed simultaneously for more than two variables.
167
For example in estimating the ultimate recovery (UR) for an oil reservoir, one would need to use the following variables: UR
=
area
. thickness
N . -
.
G
1
(I.
so. - .
RF
BO
the undefined variables so far in the text are: (I
porosity
So
the oil saturation in the pore space
Bo the formation volume factor of the oil (rb/stb); linked to the shrinkage of oil as it is brought from the subsurface to the surface RF recovery factor; the recoverable fraction of oil initially in place The Monte Carlo simulation is generating a limited number of possible combinations of the variables which approximates a distribution of all possible combinations. The more sets of combinations are made, the closer the Monte Carlo result will be to the theoretical result of using every possible combination. Using “Crystal Ball”, one can watch the distribution being constructed as the simulation progresses. When the shape ceases to change significantly, the simulation can be halted. Of course, one must remember that the result is only a combination of the ranges of input variables defined by the user; the actual outcome could still lie outside the simulation result if the input variable ranges are constrained. If two variables are dependent, the value chosen in the simulation for the dependent variable can be linked to the randomly selected value of the first variable using the defined correlation.
A Monte Carlo simulation is fast to perform on a computer, and the presentation of the results is attractive. However, one cannot guarantee that the outcome of a Monte Carlo simulation run twice with the same input variables will yield exactly the same output, making the result less auditable. The more simulation runs performed, the less of a problem this becomes. The simulation as described does not indicate which of the input variables the result is most sensitive to, but one of the routines in “Crystal Ball and @Risk” does allow a sensitivity analysis to be performed as the simulation is run.This is done by calculating the correlation coefficient of each input variable with the outcome (for example between area and UR). The higher the coefficient, the stronger the dependence between the input variable and the outcome.
168
r Random Number Generator
x1
/’ /
x2,/ /
\\\X3
\
\\
x4
cumulative
value
value
Figure 6.11 Schematic of Monte Carlo simulation
6.2.5 The parametric method The parametric method is an established statistical technique used for combining variables containing uncertainties, and has been advocated for use within the oil and gas industry as an alternative to Monte Carlo simulation. The main advantages of the method are its simplicity and its ability to identify the sensitivity of the result to the input variables. This allows a ranking of the variables in terms of their impact on the uncertainty of the result, and hence indicates where effort should be directed to better understand or manage the key variables in order to intervene to mitigate downside and/or take advantage of upside in the outcome. The method allows variables to be added or multiplied using basic statistical rules, and can be applied to dependent as well as independent variables. If input distributions can be represented by a mean, and standard deviation then the following rules are applicable for independent variables:
Sums
(say ci = ai + bi, where ai and bi are distributions)
1
the sum of the distributions tends towards a Normal distribution
2
the mean of the sum of distributions is the sum of the means;
Pc
=
Pa
+
l.4)
169
3
the variance of the sum of distributions is the sum of the variances;
Products (say ci = ai bi, where ai and bi are distributions) 4
the product of the distributions tends towards a Log-Normal distribution
5
the mean of the product of distributions is the product of the means
I-1, =
* kb
For the final rule, another parameter, K, the coefficient of variation, is introduced,
K
6
=
5
P
the value of (1 + K2) for the product is the product of the individual (1 + K2) values (1 +K);
= (1 +K):
(1 +Kb2)
Having defined some of the statistical rules, we can refer back to our example of estimating ultimate recovery (UR) for an oil field development. Recall that
U R = area From the probability distributions for each of the variables on the right hand side, the values of K, p, 5 can be calculated. Assuming that the variables are independent, they can now be combined using the above rules to calculate K, p, 5 for ultimate recovery. Assuming the distribution for U R is Log-Normal, the value of UR for any confidence level can be calculated. This whole process can be performed on paper, or quickly written on a spreadsheet. The results are often within 10% of those generated by Monte Carlo simulation. One significant feature of the Parametric Method is that it indicates, through the (1 + Ki2) value, the relative contribution of each variable to the uncertainty in the result. Subscript i refers to any individual variable. (1 + Ki2)will be greater than 1.O; the higher the value, the more the variable contributesto the uncertainty in the result. In the following example, we can rank the variables in terms of their impact on the uncertainty in UR. We could also calculate the relative contribution to uncertainty.
170
Figure 6.12 Ranking of impact of variables on uncertainty in reserves The purpose of this exercise is to identify what parameters need to be further investigated if the current range of uncertainty in reserves is too great to commit to a development. In this example, the engineer may recommend more appraisal wells or better definition seismic to reduce the uncertainty in the reservoir area and the net-to-gross ratio, plus a more detailed study of the development mechanism to refine the understanding of the recovery factor. Afluid properties study to reduce uncertainty in Bo(linkedto the shrinkage of oil) would have little impact on reducing the uncertainty in reserves. This approach can thus be used for planning data gathering activities planning how to mitigate the effects of downside in key variables planning how to take advantage of upside in key variables
6.2.6 Three point estimates : a short cut method If there is insufficient data to describe a continuous probability distribution for a variable (as with the area of a field in an earlier example), we may be able to make a subjective estimate of high, medium and low values. If those are chosen using the p85, p50, p15 cumulative probabilities described in Section 6.2.2, then the implication is that the three values are equally likely, and therefore each has a probability of occurrence of 1/3. Note that the low and high values are not the minimum and maximum values.
171
To estimate the product of the two variables below, a short cut method is to multiply the low, medium and high values in a matrix (in which numbers have been selected).
Figure 6.1 3 Combining three point estimates Note that the low value of the combination is not the absolute minimum (which would be 4, and is still a possible outcome), just as the high value is not the maximum. The three values (which are calculated by taking the mean of the three lowest values in the matrix etc.) represent equally likely outcomes of the product A*B, each with a probability of occurrence of 1/3. This short-cut method could be repeated to include another variable, and could therefore be an alternative to the previous two methods introduced. This method can always be used as a last resort, but beware that the range of uncertainty narrows each time the process is repeated because the tails of the input variables are always neglected. This can lead to a false impression of the range of uncertainty in the final result.
